Tuesday, May 22, 2012

"Lean Networking" and the Utility advantage


Martin Geddes, one of the brighter guys in a field full of smart people, writes intelligently and simply about what a modern telco company need to do to succeed. It's not easy to write both intelligently and simply about anything, especially telecommunications business models, but Geddes achieves that in an essay entitled "Lean Networking."

The basic trope he uses is to describe modern telecommunications companies as crucially similar to classic, massive, industrial manufacturing plants—they are centrally concerned with keeping the expensive plant running at full capacity all the time. They want to keep their networks full all the time and central to that is not overbuilding and viewing both network queues and QoS devices as fundamentally a way of keeping the pipes full. Quality, in guise of loss and delay, is is conceived of as something that happens after you've assured basic network efficiency, in the guise of full pipes. (You may now have an inkling of the underlying conundrum that AT&T' has created for itself.)

In contrast modern, "lean," manufactories are focused on what constrains the production of quality goods and in eliminating those constraints. Quality, conceived of as low defect and high user satisfaction, is the first focus of lean manufacturing. Efficiency is something that you work up to once you've got the quality where it needs to be to satisfy your market. So keeping the factory running at capacity at all times is not the first priority and is achieved by ruthlessly eliminating any constraints on supplying quality goods as they are needed even if that means plant downtime as the demand ebbs periodically. Maintaining some excess capacity is part of the price of viewing manufacturing in this way. By analogy, the lean network of the title of the essay is one which views minimizing loss and delay by focusing on those parts of the network that constrain the network—chiefly contention—eliminating those, and dealing with ebbs and flows by network planning allows excess capacity in accordance with anticipated large flows.

Geddes take-home idea is that the only way to have both efficiency and quality is to ensure quality first. Focusing on efficiency first results in the big, bulky, plants that have to run at full capacity all the time. And that sort of full capacity can only be achieved by methods which compromise on quality. But if you go for quality first you can adapt to the ebbs and flows exactly by minimizing queuing and other tactics which actually ensure contention. Only the quality-first route can take you to the very most efficient network.

Ok, you may say, but why talk about something so arcane in a blog focused on a local municipal network. Well...

Lafayette—and other fiber municipal or small networks.  
Geddes is focused on advising enormous,vertically integrated creatures like AT&T. As he notes the chief constraint on quality, contention, is worse at the local edges of the network. The solution is to both being willing to refuse to allow new entrants at peak times and to build more capacity to reduce the frequency of such moments. (Like lean manufactories.) The old telephone network that AT&T and its brethren built worked this way: like most utilities it was luxuriously over built and, in those rare moments of contention, new calls were simply refused; you were told "all circuits are busy" on those rare occasions when capacity was challenged. Interestingly, cable companies had their own version of this: they devoted discrete capacity to each channel, classically, each channels full info flowed past your tv's receiver all the time. It was hugely wasteful in terms of efficiency since you could only watch one channel at a time.

The advent of Internet Protocol (IP) and associated protocols changed all that; the key term is "best effort"—IP networks are not "all or nothing" networks like the old phone service or the classic cable model; instead they can trade quality for capacity. It's more efficient overall. And there is a strong engineering and financial push in the direction of efficiency; the two mindsets are in agreement on this issue.  The cost-savings occurs at the very edges where Geddes notes there is the most contention. This is at least partly because it is the most expensive part of the network in which to upgrade capacity. It's where the compromises are most profitably made. Old technologies that are more expensive in the long run are given spotty and gimmicky upgrades to extend their life and only the relatively inexpensive core lines are given the newer, more capable, and more easily upgraded fiber treatment.

That (finally!) is where entities like LUS Fiber come in. If the big guys can't see their way toward fixing the edges local folks can...and the end result can—if LUS focuses its bandwidth purchases on ensuring contention-free links at the regional level—result in a user network like the one Geddes would recommend: a network which is built to minimize contention first, and build efficiency on top of that.

It's a utility mentality. And with its LUS Fiber utility Lafayette is a "back to future" sort of place where the local supplier of the most constrained part of the network returns to viewing service as a quality issue and is willing to oversupply capacity in order to get quality...which, if Geddes take-home idea is right, is the only way to get to real quality and maximum efficiency.

LUS' biggest advantage in the modern business environment may well be its oldest attribute: a commitment to running the network like a utility.

No comments: